United States
Environmental Protection
Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711'
Research and Development
EPA/600/S3-86/052 Dec. 1986
&EPA Project Summary
Local and Regional
Contributions to Urban
Particulate Matter
S. A. Batterman, J. A. Fay, and D. Golomb
This report summarizes the data
analysis of two extensive field studies
on urban particulate matter: the 1974-
77 St. Louis (RAPS) and the July/August
1982 Philadelphia (PAFS) studies. Spe-
cial emphasis is placed on the "dichot"
measurements of particles that segre-
gated the fine fraction (<2.5 jim) and
the coarse fraction (2.5 - 15 p,m in St.
Louis and 2.5 - 10fim in Philadelphia).
The major conclusion of this study is
that in both cities the majority (more
than 50%) of the total mass collected by
the dichots is of regional, not local,
origin. The regional share is about
equally large for long-term (yearly, sea-
sonal, monthly) and short-term (24-
hours) averaging of concentrations. In
the fine fraction, the regional compo-
nent is even larger, 60% in St. Louis and
83% in Philadelphia. This conclusion is
supported primarily by the observation
that with the exception of a single site
in both St. Louis and Philadelphia, all
other sites within the extensive moni-
toring network, including remote rural
sites, show a very low gradient of con-
centrations. This observation shows
that the majority of collected aerosols
do not originate from local sources, but
must come from distant, regional
sources that impact all monitors
equally.
The annual average composition of
PM-15 in St. Louis is 21% sulfate as S04,
39% crustal matter (SiO2, AI2O3, Fe2O3,
CaO, MgO), 35% "unknown" (i.e., not
determined by the routine analytical
method. X-ray fluorescence), and 5%
others (mainly trace metals). The sum-
mer monthly average composition of
PM-10 in Philadelphia is 34% SO* 14%
crustal, 51% unknown, and 1% others.
The unknown contains carbonaceous
matter (elemental carbon and con-
densed organic matter, ammonium, ni-
trate, and water). Peak 24-hour compo-
sitions are not greatly different from
the above.
Given the large proportion of re-
gional contribution to and the chemical
makeup of PM-10(15), neither disper-
sion nor receptor modeling based on
local emission inventories and elemen-
tal composition is likely to accurately
predict or interpret particle levels in
urban airsheds.
This Project Summary was devel-
oped by EPA's Atmospheric Sciences
Research Laboratory, Research Triangle
Park, NC, to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering infor-
mation at back).
Introduction
This report summarizes the findings
of a two-year research effort on inhal-
able particle characteristics in urban air-
sheds. The research is based on data of
two extensive urban field programs
conducted by EPA in St. Louis (1974-
1977) and in Philadelphia (1982). The St.
Louis Regional Air Pollution Study
(RAPS) was probably the largest moni-
toring effort ever undertaken to charac-
terize the temporal and spatial charac-
teristics of air quality in an urban
environment, with special emphasis on
inhalable particles. Among other moni-
toring devices, the RAPS campaign op-
-------
erated 10 dichotomous inhalable parti-
cle samplers continuously for over
three years, covering an area of 30 x 90
km. In this study, one year of data (1976)
was analyzed. The Philadelphia Aerosol
Field Study (PAFS) was of shorter dura-
tion, four weeks in July/August 1982,
and of smaller spatial extent—six di-
chotomous samplers covering an area
of 20 x 30 km.
This report provides an in-depth dis-
cussion and analysis of the two data
bases, including determinations of par-
ticle composition, temporal and spatial
characteristics, and meteorological in-
fluences. The data are then analyzed to
separate the components of particulate
pollution concentrations that are 'at-
tributable to sources within or outside
the local urban area. These are referred
to as local and regional components,
and the effort helps define the amount
of air quality improvement that can be
obtained from local emission controls.
Next, dispersion and receptor (statisti-
cal source apportionment) models are
applied to the data sets to evaluate
model performance individually and in
comparison. Finally, the report dis-
cusses the concept of combining dis-
persion and receptor modeling ap-
proaches in a mixed model that could
improve predictive and analytic capabil-
ities geared toward recommendation of
control strategies.
Ambient Particle Data
High PM-10 and PM-15 levels gener-
ally result from large contributions of
both fine and coarse size fractions.
While fine and coarse fractions are
about equal in St. Louis, in Philadelphia
the fine fraction is dominant at most
sites. Based on the different composi-
tions and low correlations, the two size
fractions appear to have largely sepa-
rate origins.
About one-half of the fine fraction
particles and one-third of PM-10 consist
of S04. SO4 levels are highest in the
summer, probably due to the faster oxi-
dation of S02 to SO4 associated with in-
creased photochemical activity, high
temperatures and high humidities. At
most urban and rural sites, concentra-
tions of fine fraction mass and S04 are
similar and their pollutant roses are
nearly identical; both imply the impor-
tance of regional sources. At high con-
centrations, the fraction of trace metals,
indicative of local sources, increases in
St. Louis but decreases somewhat in
Philadelphia, possibly due to the domi-
nance of regional SO4 in the latter city.
Coarse fraction particles contain large
amounts of Al, Si, Ca, Fe and Mn and
appear to be of crustal origin. This
crustal material composes a relatively
constant fraction of the aerosol. On av-
erage, coarse fraction levels in Philadel-
phia are much lower than in St. Louis
(summer averages of 9 vs. 25 ng/m3),
probably due to the change in sampler
design, which excludes larger particles.
Source differences between the two cit-
ies may also be important. The highest
coarse fraction concentrations occur
during dry and dusty periods; concen-
trations tend to decrease during pro-
longed or heavy precipitation.
Local and Regional
Components of Particles
Concentrations
Because particles are formed and
transported over long distances, both
local and regional or distant emission
sources contribute to ambient particle
concentrations. Thus, particle concen-
trations may be viewed as the sum of
contributions from local and regional
emission sources. The regional compo-
nent of the total particle mass collected
at a receptor is termed "background."
Hereafter, "regional component" and
"background" are used interchange-
ably. Local sources, situated within the
airshed, produce concentration levels
which generally increase toward the
sources. This is the "urban increment."
The regional or background component
arises from the long-range transport of
pollutants into the airshed and attains
about equal levels at all locations within
the airshed.
It is assumed that the upwind or re-
gional monitoring site receives the low-
est concentration in the monitoring net-
work. In both rural and urban areas, the
local increment or contribution is the
difference between the highest and low-
est concentrations.
The annual average concentrations at
the St. Louis sites of the PM-15, fine and
coarse fractions, were calculated and
the background share averages 57.5%
of the PM-15 at the 10 sites. The fine
fraction background is 60%; the coarse
fraction background is 55%. The ratio of
average concentrations at the central in-
dustrial site to the most outlying site is
only about 2 for PM-15, 1.75 for fine
fraction, and 2.3 for coarse. There are
two significant conclusions to be
drawn. Concentrations at rural sites are
due almost entirely to background parti-
cles and therefore receive little contri-
bution from the metropolitan St. Louis
sources. Equally important, the back-
ground concentration is a significant
fraction of the total concentration at
each site, even for the central city sites.
The interesting conclusion is that on
low pollution days local sources may
contribute more to the total concentra-
tions. On high pollution days (when
standards are likely to be exceeded),
local impacts become less pronounced
and most particles seem to come from
outside the network.
The monthly average concentrations
were considered at the six Philadelphia
sites. Site averaged concentration are
computed for five sites only, excluding
one with peculiar results. The back-
ground share of the average PM-10 con-
centrations is 77%. The fine fraction
background is 83% and the coarse 64%.
The ratio of concentrations at the center
city site to that at the most rural site is
less than 1.2 for total mass and fine frac-
tion, and 1.6 for the coarse fraction.
Thus, in Philadelphia the background
share is even larger than in St. Louis,
and the relative difference between city
center and outskirts is smaller. Looking
at daily average concentrations at sites
in Philadelphia, the ratio to background
is usually less than two. As in St. Louis,
the range is greater at lower overall con-
centrations. This again indicates that
the proportionate contribution of local
sources is less at high concentrations
when exceedances might occur.
In summary, the regional component
composed well over half of the average
fine fraction levels at most monitors in
both cities and about half of the coarse
fraction concentrations. The back-
ground fraction is relatively constant;
thus, local and regional levels appear to
change together.
Dispersion Modeling
A long-term version of the Particle
Episodic Model (PEM) is applied to ap-
portion and predict particle concentra-
tions in St. Louis and Philadelphia. In
the evaluation, only the local contribu-
tions are modeled. The observed local
component is obtained by subtracting
the regional component as described in
the preceding section.
Estimated Philadelphia fine fraction
emissions are about twice the coarse
fraction, whereas in St. Louis the coarse
fraction emissions are estimated to be
higher. These estimates may in part ex-
plain the observed differences of ambi-
ent particle levels in the two cities. Area
sources are the primary contributors of
-------
particles. Host cells, the area source cell
in which the receptor is located, account
for about half of the total particle predic-
tion. Point sources produce roughly half
of the S02 and S04. Generally, a few
sources provide most of the contribu-
tions at the receptors.
In St. Louis, however, particle levels
are overpredicted, especially in the
coarse fraction, and predictions do not
have the correct spatial distribution.
Most likely, the St. Louis emission in-
ventory does not accurately reflect PM-
15 emissions. In St. Louis, fine and
coarse particle emissions were derived
from TSP data, which in many cases
may not be related to PM-15 emissions
but rather to fugitive dust. At short aver-
aging times* model predictions have
low correlation (0.2-0.3) with observa-
tions. In Philadelphia, long-term
predictions are in reasonable agree-
ment with measurements. The mean
levels and the spatial variation at most
sites are quite well modeled.
Statistical Models for Source
Apportionment
Receptor methods are useful only for
the apportionment of those sources that
have distinct composition and are pre-
dominantly of local origin, e.g., vehi-
cles, incinerators, metal processing in-
dustries, and some oil-related sources.
Receptor models do not separate local
from regional sources that have similar
elemental composition.
According to the multiple linear re-
gression (MLR), 56% of the fine fraction
and 34% of the total PM-15 levels in St.
.ouis are because of S04. Previous
studies using summer data only, at-
ributed a greater percentage to SO4 (59
o 84% of the fine fraction). This could
>e due partially to higher transforma-
ion rates in the summer. Apportion-
nents of other sources are similar to the
>revious studies. Crustal sources ac-
:ount for 85% of the coarse fraction and
16% of PM-15. Other sources tentatively
dentified include road salting, Indus-
rial emissions and incineration. The
/ILR model explains 50-90% of the vari-
mce of particle concentrations and indi-
ates that S, Ca and Cu are stable trac-
rs.
In Philadelphia, the MLR approach in-
licates that SO4 accounted for 52-65%
if fine concentrations and 39-52% of
'M-10 concentrations. Crustal compo-
ents account for 30-50% of the coarse
•action; oil and/or refinery sources (V)
ccount for 2-11% of PM-10 concentra-
ons; vehicular sources may contribute
8-15% of PM-10. The identification of
other factors is more speculative. Incin-
eration (Cu) may'account for 5-8% of
fine fraction concentrations; and crustal
or fertilizer sources (P, K) may compose
6-25% of particle concentrations.
MLR results indicate that the mass
loading factor of elemental S is 5.1 in St.
Louis and 4.1 in Philadelphia. The stoi-
chiometric factor for (NH4)2S04 is 4.1;
thus, in St. Louis, the sulfur components
may have carried additional matter, per-
haps water.
Mixed Models
With a few exceptions, statistical (re-
ceptor) and physical simulation (disper-
sion) models remain separate ap-
proaches in air quality modeling. To
provide more accurate results, a "mixed
model" was developed that incorpo-
rates features of both dispersion and re-
ceptor modeling.
In some respects, the mixed model
described here resembles the "inverse"
dispersion model and state-space ap-
proaches. In the inverse dispersion
model problem, source emissions
(rather than ambient concentrations)
are predicted using ambient observa-
tions. The mixed model differs in that
multiple pollutants, source, composi-
tions and prior information are consid-
ered.
The model consists of several compo-
nents. First, a dispersion model is used
to calculate transfer coefficients. These
coefficients indicate the contribution of
local emission sources to receptors for
the meteorological conditions during
the sampling period. The Particle
Episodic Model (PEM) is used with opti-
mized dispersion parameters. Second,
predictions of mass and elemental con-
centrations at receptor sites are made
using the transfer coefficients and prior
information. The latter includes the ele-
mental composition and the particle
emission rate of sources. Spatial and
temporal aggregation is used to reduce
the number of unknowns and simplify
the estimation problem. Third, the prior
information—emission rates and
source compositions—is revised so that
predicted mass and elemental concen-
trations best correspond to ambient
measurements. Measurements col-
lected at all receptor sites are used. Lin-
ear Bayesian estimates are used to cor-
rect for missing data. As the
distributions of the parameters are un-
known, a parametric approach is used
in which ranges of uncertainties and co-
variances are selected. The primary out-
puts of the model are posterior esti-
mates of emission rates and elemental
compositions. Apportionments are
derived as the product of the estimated
emission rates and the transfer coeffi-
cients.
Mixed models may be used in many
applications. First these models can rec-
oncile different apportionments gener-
ated by simulation and receptor mod-
els. Second, the results provide a check
on the accuracy of the source inventory.
Third, the appropriateness of source
compositions may be assessed. Fourth,
it may be possible to identify unknown
sources or detect locations of accidental
releases. Fifth, the approach forces the
explicit quantification of uncertainty.
Conclusions
While the two field studies were dif-
ferent in scope and extent, and even
used different instrumentation and
schedules, the main conclusion is the
same: the majority (more than 50% of
the inhalable particle mass, whether av-
eraged over a day, month, or year is not
attributable to sources that lie within the
metropolitan city limits, but is probably
because of regional sources.
Other conclusions regarding inhal-
able particles in St. Louis and Philadel-
phia follow:
• In St. Louis, the fine fraction (less
than 2.5 ^.m diameter) and coarse
fraction (2.5-15 jtm) masses are
about equal; in Philadelphia the
fine fraction (<2.5 (im) and coarse
fraction (2.5-10 p-m) masses are
about 3:1.
• In St. Louis, about 40% of the fine
fraction consists of sulfate (SO4);
this percentage is even larger if it is
assumed that most of the sulfate
consists of hydra ted ammonium
sulfate. In Philadelphia, 43% of the
fine fraction is SO4. The coarse frac-
tions are dominated by crustal
components.
• In St. Louis, 24-hourly PM-15 con-
centrations exceeded 150 n.g/m3
several times, which is the pro-
posed lower range of the 24-hour
PM-10 standard for particles. In
Philadelphia, during the four weeks
of monitoring, this range was ap-
proached only once at one site.
• The ratio of annual network aver-
ages of inhalable particle mass to
total suspended particle mass is
about 0.5; however, the ratios of
24-hour averages may have a wider
range, from 0.25 to 0.75.
-------
Dispersion modeling, at best, can
account only for the local incre-
ment of inhalable particles; not the
background that, as indicated
above, appears to be of regional
origin. Dispersion modeling re-
quires an accurate emission inven-
tory of inhalable particles (and their
gaseous precursors), which cannot
be obtained simply by assuming
that there is a constant ratio of
emissions of IP/TSP.
Receptor modeling is only useful
for the apportionment of those
sources that have distinct composi-
tions and are of local origin, e.g.,
vehicles, incinerators, metal proc-
essing industries, and oil combus-
tion-refining. Many sources have
similar profiles and cannot be sepa-
rated by receptor methods.
The mass loading factor of elemen-
tal sulfur (sulfur-related mass di-
vided by sulfur mass) is about five
in St. Louis and about four in
Philadelphia. This factor is larger
than the stoichiometric factor of
three for S04, indicating that ele-
mental sulfur carries along some
other species, probably ammonium
and water.
A "mixed" dispersion-receptor
model was developed in which
multiple pollutants, source compo-
sitions, and dispersion model-
derived transfer coefficients were
used. The mixed model is computa-
tionally very intensive but yielded
some useful information regarding
emissions.
S. A. Batterman, J. A. Fay, andD. Golomb are with the Massachusetts Institute of
Technology. Cambridge. MA 02139.
Jack H. Shroffler is the EPA Project Officer (see below).
The complete report, entitled "Local and Regional Contributions to Urban
Paniculate Matter." (Order No. PB 86-236 965/AS; Cost: $11.95. subject to
change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park. NC 27711
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S3-86/052
/'.- ;:~0.i:£-
0000329 PS
230 S DEARBORN STREET
CHICAGO It 60604
------- |